Study On The Processing Method And Application Of Visual Information In Rock-Filled Concrete Construction Management

Rock-Filled Concrete (RFC) has been employed in a number of hydraulic engineering as a new type of concrete since it was developed in 2003. RFC construction requires quality control of proper rock size distribution and qualified self-compacting concrete (SCC). By now, the quality control of RFC construction is conducted by traditional methods of collecting data using human judgement with low precision. In this study, a non-contact, objective and convenient system is proposed and developed based on automatical visual information processing methods for improving the quality management on RFC construction sites.RFC dam construction projects are used for the development of this visual information processing system. The practical demands of construction site management are analyzed, and classified into two parts: multi-moving-targets tracking and static objects detection.For multi-moving-targets, the track of moving targets such as vehicles, hoppers and SCC material on construction sites are required for management. Using a fixed video camera for image collecting, the main difficulties for multi-moving-targets tracking are the changing background and the presence of occluded targets in the complex environment of the construction site. The existing methods for targets detection, location and tracking under occlusions are analyzed. The adaptive background model based on most frequent value is proposed for targets detection. The connected component labeling algorithm using run recursive method is improved for targets location. The matching method according to judgment of consecutive trajectory is developed for targets tracking. The experimental results show that the moving targets can be real-timely detected,located and tracked precisely using these proposed algorithms. For static objects detection, the characteristic of objects needed to be detected for quality management is size distribution of rock mass on construction sites. The main difficulty for detection is to segment the overlapped objects. The existing methods for separating overlapped objects are reviewed, and watershed segmentation method is adopted to detect the size of each rock from an image. The experimental results compared with manual detection show that the method is feasible and effective.The visual information processing system is developed based on the proposed algorithms and methods. This system is applied to track the hopper for evaluating the production of SCC, and to detect the rock size distribution for quality control. In the future, the system can be developed for visual information processing of other targets in Rock-filled concrete construction management, or even for targeting management in other type of construction process.